The impact of chatGPT and other large language models on physics research and education

Friday 21 Jul 2023, 08:55 → 17:20 America/New_York

32-123 (MIT)

Kevin Burdge (MIT), Joshua Borrow (Massachusetts Institute of Technology), Mark Vogelsberger (MIT)

Join us for a one-day meeting within the MIT Physics Department to explore the growing significance of Language Models (LLMs). This event will feature a series of contributed talks from esteemed members of the department, along with informative overview presentations. We will delve into the practical applications of LLMs in various domains, including Teaching & Administration, Research & Writing, Programming & Data Analysis, and the underlying technology of LLMs. This gathering offers an excellent opportunity to learn, share insights, and foster collaboration within the MIT community. The event is free, and registration is open to all members of the MIT community.

The event will be hybrid, and can be attended on Zoom. The link will be sent out to registered participants before the event, and through the MIT Physics e-mail lists. If you want access the the link and have missed the registration deadline, please contact Josh Borrow (borrowj@mit.edu), listing your name, title, affiliation, and .edu e-mail.

Written by ChatGPT-3.5 (Prompt: I am organising a meeting within the MIT physics department about LLMs. The meeting will allow for contributed talks from various members of the department, and some overview talks from others. It will be a one day event. We will discuss how LLMs are relevant to Teaching & Administration, Research & Writing, Programming & Data Analysis, as well as the technology of LLMs themselves. Please write me a paragraph or two for the description of the event, mentioning it is free and registration is open to the entire MIT community. Please avoid flowery language.)

Video recordings of workshop

Zoom link

09:00 → 11:50 The computer science underlying large language models

This session will focus on a discussion of the computer science underlying large language models.

Convener: Kevin Burdge (MIT)

09:00 Introductory remarks

Speaker: Dr Kevin Burdge (MIT)

09:05 Keynote address: Towards Machines that can Learn, Reason, and Plan (remote)

Speaker: Prof. Yann LeCun (NYU/Meta)

10:00 Generative AI gives some spectacular demonstrations, but I don’t know how!

Speaker: Prof. Peter Szolovits (MIT)

10:15 Language Models as World Models?

Speaker: Prof. Jacob Andreas (MIT)

10:30 Coffee Break

10:45 Some experiments with LLMs, or why the world you were raised to survive in no longer exists

Speaker: Dr Siddharth Mishra-Sharma (MIT/IAIFI)

11:00 The Quantization Model of Neural Scaling

Speaker: Eric Michaud (MIT)

11:15 Keeping AI under control through mechanistic interpretability

Speaker: Prof. Max Tegmark (MIT)

11:30 Open discussion on broader societal impacts of LLMs, and implications for physics majors

Speakers: Dr Kevin Burdge (MIT), Dr Joshua Borrow (MIT)

11:50 → 12:40 Lunch

12:40 → 14:30 The use of large language models in research

12:40 Collaborating with ChatGPT to train an unsupervised machine learning algorithm on astronomical data

Speaker: Emma Chickles (MIT)

12:55 Fine-Tuning LLMs for Understanding Fundamental Physics Concepts and Problems

Speaker: Rumen Dangovski (MIT)

13:10 Building a platform for understanding scientific ideas by offloading increasing amounts of the research process to AI

Speaker: Joshua Ramette (MIT)

13:25 Using natural language processing-based information to aid the detector characterization and commissioning activities at various gravitational wave observatories

Speaker: Dr Nikhil Mukund (MIT)

13:40 Interactive demo session on using chatGPT in teaching and research-type applications

Speaker: Dr Kevin Burdge (MIT)

13:50 Open discussion on use of LLMs in teaching and research

Speakers: Dr Joshua Borrow (MIT), Dr Kevin Burdge (MIT)

14:10 Coffee break

14:30 → 17:20 The use of large language models in teaching/administration

14:30 How LLMs Are Changing Teaching at MIT

Speaker: Prof. Christopher Capozzola (MIT)

14:45 Why I’m excited (but nervous) about using GPT in my MIT teaching

Speaker: Prof. Daniel Jackson (MIT)

15:00 Combing AI transformers for convincing character generation

Speaker: Dr Loyd Hoyt Waites (MIT)

15:15 AI Augmented Support for Computing Services

Speaker: Julius Heitkoetter (MIT)

15:30 Answering student questions all-at-once! A case study for AI in the classroom

Speaker: Dr Louis Deslauriers (Harvard)

15:45 Learning the other way: on AI and acing MIT

Speaker: David Koplow (MIT)

16:00 Conversation on the Use of ChatGPT in Fall 2023 Physics Teaching

Speaker: Dr Michelle Tomasik (MIT)

16:10 Capstone talk: LLMs for Physics, and Physics for LLMs

Speaker: Dr Stephen Wolfram (Wolfram Research)

17:10 Closing remarks

Speaker: Dr Kevin Burdge (MIT)